Pyspark：检查列值是否单调递增

Question

问题： 给定下面的 pyspark 数据框，是否可以按行检查“some_value”是否确实增加了（与前一行相比）使用 window 函数（参见下面的示例）?

没有滞后的解决方案是首选，因为我将有多个列，如“some_value”，我事先不知道有多少以及它们的明确名称。

示例： 在这里，我想获得像“FLAG_INCREASE”这样的列。

+---+----------+---+----------+
| id|     datum|lfd|some_value| FLAG_INCREASE
+---+----------+---+----------+ ------------+
|  1|2015-01-01|  4|      20.0| 0
|  1|2015-01-06|  3|      10.0| 0
|  1|2015-01-07|  2|      25.0| 1
|  1|2015-01-12|  1|      30.0| 1
|  2|2015-01-01|  4|       5.0| 0
|  2|2015-01-06|  3|      30.0| 1
|  2|2015-01-12|  1|      20.0| 0
+---+----------+---+----------+--------------+

代码：

import pyspark.sql.functions as F
from pyspark.sql.window import Window
from pyspark.sql import Row
 
 
row = Row("id", "datum", "lfd", "some_value", "some_value2")
df = spark.sparkContext.parallelize([
    row(1, "2015-01-01", 4, 20.0, 20.0),
    row(1, "2015-01-06", 3, 10.0, 20.0),
    row(1, "2015-01-07", 2, 25.0, 20.0),
    row(1, "2015-01-12", 1, 30.0, 20.0),
    row(2, "2015-01-01", 4, 5.0, 20.0),
    row(2, "2015-01-06", 3, 30.0, 20.0),
    row(2, "2015-01-12", 1, 20.0, 20.0)
]).toDF().withColumn("datum", F.col("datum").cast("date"))
 
+---+----------+---+----------+
| id|     datum|lfd|some_value|
+---+----------+---+----------+
|  1|2015-01-01|  4|      20.0|
|  1|2015-01-06|  3|      10.0|
|  1|2015-01-07|  2|      25.0|
|  1|2015-01-12|  1|      30.0|
|  2|2015-01-01|  4|       5.0|
|  2|2015-01-06|  3|      30.0|
|  2|2015-01-12|  1|      20.0|
+---+----------+---+----------+

Answer 1

你只需要 lag:

from pyspark.sql import functions as F, Window

df = df.withColumn(
    "FLAG_INCREASE",
    F.when(
        F.col("some_value")
        > F.lag("some_value").over(Window.partitionBy("id").orderBy("datum")),
        1,
    ).otherwise(0),
)

Pyspark：检查列值是否单调递增

Pyspark: Check if column values are monotonically increasing

python

data-processing

apache-spark

pyspark